HAL Id: hal-01533112
https://hal-amu.archives-ouvertes.fr/hal-01533112
Submitted on 5 Feb 2019HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Remote Sensing of Industrial Palm Groves in Cameroon
Prune Christobelle Komba Mayossa, Sébastien Gadal, Roda Jean-Marc
To cite this version:
Prune Christobelle Komba Mayossa, Sébastien Gadal, Roda Jean-Marc. Remote Sensing of Industrial Palm Groves in Cameroon. ASM Science Journal, Akademi Sains Malaysia, 2017, 10 (1), pp.16-45. �hal-01533112�
Remote Sensing of Industrial Palm Groves
in Cameroon
Prune Christobelle Komba Mayossa1,2*, Sébastien Gadal1 and Jean-Marc Roda2,3
The measurement of biomass can be obtainedfrom remote sensing analysis and modelling , the impacts of which are related to oil palm cultivation in industrial plantations. Our study aims at producing a spatial model for oil palm biomass estimation, at different scales of spatial analysis. The study was carried out in the industrial plantations of the Cameroonian Society of Palm Groves (SOCAPALM). The developed methodology combined: (i) the mapping of palm groves (Kumar, 2015), (ii) the characterisationof palm groves (Gadal, 2013), (iii) biomass estimation, and (iv) the comparison of the obtained results with Spot6, Landsat 7 ETM+ and Landsat 8 OLI images from 2001 to 2015. The first results were obtained for the mapping of the SOCAPALM industrial palm groves between 2001 and 2015. The obtained maps were highly correlated (Kappa of 0.91 for Spot 6, 0.92 for Landsat 7 and 0.82 for Lansat8), however, because of the presence of mixed pixels, some confusion between oil palm and other classes were observed. One of the factors affecting biomass estimation is spatial accuracy. Several improvements have been suggested : (1) mapping palm groves at a subpixel scale using super-resolution mapping; (2) developing a classification system of cartographic products. The use of satellites images with different spatial resolutions may also help to generate new data taking into account the level of spatial analysis.
Key words: Palm groves, biomass, remote sensing, spatial accuracy, energy, modelling
IntroductIon
In Cameroon, oil palm gainedgreat economic importance along with industrializstion dating from the colonial period (Elong, 2003). The income generated from oil palm cultivation have developed agro-industries such as SOCAPALM (Cameroonian Society of Palm groves). These activities havehigh-yields but low costs of returns (Rival, 2013) andhave also causedsocio-environmental damages including deforestation, loss of biodiversity, pollution, etc. The measurement of biomass can be obtained from remote sensing analysis and modeling, the impacts of which are related to oil palm cultivation in industrial plantations. Biomass is an important variable in many ecological, environmental and agricultural studies. It is a renewable source of energy used to monitor, and quantify agricultural production.
In the oil palm industry, remote sensing has become the main source of biomass estimation (Lu, 2005). Using remote sensing, several studies have been developed to monitor the production and yields of oil palm. However, thesestudies only focused on palm grove expansion (Thenkabail, 2004), age assessment of palm groves(Chemura
et al. 2015) and the conversion of forest areas into oil palm cultivation (Morel et al. 2011).
1Aix-Marseille Université, CNRS ESPACE UMR 7300, Europôle méditerranéen de l’Arbois - BP80 - 13545 Aix-en-Provence
Cedex 4, France
2French Agricultural Research Centre for International Development (CIRAD), International Research Unit Biomass, Wood,
Energy, Bioproducts (UPR BioWooEB), F-34398 Montpellier, France
3Institute of Tropical Forestry and Forest Product (INTROP), Laboratory of Sustainable Bioresource Management (BIOREM),
Universiti Putra Malaysia 43400 UPM Serdang, Selangor (Malaysia)
Prune Christobelle Komba Mayossa et al.: Remote Sensing of Industrial Palm Groves in Cameroon
Being the first survey of its kind in the Congo basin, this work is related to thelong-term management of oil palm resources. More precisely, in the context of the opposition between biodiversity conservation and oil palm agricultural strategies, and in the REDD context,it characterises the relationship between people/agriculture and biodiversity. This work aims to:
• Produce a spatial model for oil palm biomass estimation at different scales of spatial analysis. • Establish the link between agricultural strategies of palm groves, biomass and the produced energy.
The study is focused in the industrial palm groves of the Cameroonian Society of palm groves (SOCAPALM). The largest agribusiness of oil palm in Cameroon islocated in the south west of this area.
MaterIals and Methods
Satellite images between 2001 and 2015 were used: Landsat ETM+, Landsat 8 OLI-TIRS and Spot-6 of the same season but with different acquisition dates. The methodology adopted combined the mapping of palm groves (using per-pixel classification), biomass estimation (using the combination of a linear regression model and biophysical variables) and third, the comparison of the obtained results for the earth observation platform.
results and dIscussIon
First, results were obtained for the mapping of palm groves from 2001 to 2015 (Figures 1, 2 and 3).
Figure1. Data Table1. Sensor specifications
Sensor Acquisition date Mode Resolution Cloud cover
ETM+ 26/04/2001 MS/PAN 30m/15m 30%
OLI-TIRS 26/04/2015 MS/PAN 30m/15m 30%
Figure 2. Land cover map from Landsat 7 ETM+ (2001), CNRS ESPACE UMR 7300
Prune Christobelle Komba Mayossa et al.: Remote Sensing of Industrial Palm Groves in Cameroon
Figure 4. Land cover map from Spot6 (2015) CNRS ESPACE UMR 7300.
Highly correlated maps with overall accuracies of 90% for Landsat 7, 80% for Landsat 8 and 93% for Spot 6 respectively, were obtained. However, despite the high values of Kappa (0.91 for spot 6, 0.86 for Landsat 8 and 0.92 for Landsat 7) between oil palm classes, or oil palm and vegetation classes, someconfusion was observed between forest and mature oil palms (17%); and between young and growing oil palms (5.4%).
• As some components may be common to different classes, (for example low vegetation pixels in the growing or young oil palm classes), distinct classes may share mixed pixels (Komba Mayossa, 2014).
• On the other hand, to validate the produced maps, control areas could bedigitised with Arcgis software in many classes as estimated for each image. The resulting manual classifications could then be crossed with the maximum likelihood classification result, to produce a confusion matrix and Kappa index. The ground truth plays an important role in map accuracy. The knowledge of the ground is an important key; andresults can be obtained from the validation of the digitised map and classifications. Landsat images have aresolution of 30m. The photo-interpretation is difficult especially in a heterogeneous landscape, as in the current study area, errors occur during the sampling step, which cannot take into account the within-plot heterogeneity. Thematic confusions caused by the presence of mixed pixels affects map accuracy (Chitroub, 2007). These mixed pixels result from spectral and spatial characteristics of the studied objects in the landscape, and from the methods used to map.
conclusIons
Limitations of the spatial accuracy of oil palm grove mapping is relatedto class heterogeneity, attendant mixed pixels and the method used. Several ways for improvement are possible.
• For the mapping of industrial palm groves
Firstly, per-pixel classification assumes that each pixel represents a single class only. The maximum likelyhood algorithm, used to map palm groves, ignores the mixed pixel problem. One of the solutions is to use a technique which allows mapping at the sub-pixel scale, such as super-resolution mapping (Priyaa and Sanjeevi, 2013; Muad and Foody, 2012). Secondly, as the map validation method may enhance or reduce the accuracy of the produced map, the development of a classification system of cartographic products becomes very interesting. Indeed, further studies should be focused on the evaluation of the reliability of the produced map.
• For biomass estimation
One of the factors affecting biomass estimation is the spatial resolution. The use of data with different spatial resolutions allows taking into account the issues related to the spatial accuracy of produced models. This also helps to generate new data, taking into account the level of spatial analysis.
The purpose of this work was to develop a model for oil palm biomass estimation at different scales of spatial analysis, that links agricultural strategies, biomass and produced energy for oil palm plantations in a tropical area. This mayopen prospects for the implementation of an environment observatoryfor the monitoring of agricultural areas, considering socio-economic and geographical features.
references
Elong , J.G 2003, ‘Les plantations villageoises de palmier à huile de la Socapalm dans le bas-Moungo (Cameroun): un projet mal intégré aux préoccupations des paysans’, Les Cahiers d’Outre-Mer, Revue de géographie de Bordeaux, Vol.56 , n°224, pp 401-418.
Chemura, A, van Duren, I, van Leeuwen, L. M 2015, ‘Determination of the age of oil palm from crown projection area detected from WorldView-2 multispectral remote sensing data: The case of Ejisu-Juaben district, Ghana’., ISPRS Journal of Photogrammetry and Remote Sensing, n°100, PP.118-127.
Chitroub , S 2007, ‘Annalyse des composantes indépendantes d'images multibandes: Faisabilité et perspectives’, Revue de télédétection, Vol.7, n°1-2,pp.3-4.
Gadal, S 2003, ‘Reconnaissances multi-niveaux d’unités paysagères par segmentation automatique d’images satellites’, eds Anne-Elisabeth LAQUES, in Télédétection des informations géographiques, pp.42-54.
Komba Mayossa, P.C 2014, ‘Développement d'une méthode de traitement d'images satellites pour la cartographie d'agrosystèmes à base de cocotier’, Mémoire de Master2 de l’Université de Rennes2, Rennes, France.
Lu, D 2006, ‘The potential and challenge of remote sensing-based biomass estimation’. International journal of remote sensing, vol. 27, no 7, pp. 1297-1328.
Morel, A.C, Saatchi, S.S, Malhi, Y 2011, ‘Estimating aboveground biomass in forest and oil palm plantation in Sabah, Malaysian Borneo using ALOS PALSAR data’, Forest Ecology and Management,vol. 262, no 9, pp. 1786-1798.
Muad, A.M, and Foody, G.M 2012, ‘Super-resolution mapping of lakes from imagery with a coarse spatial and fine temporal resolution’, International Journal of Applied Earth Observation and Geoinformation, vol.15, pp.79-91.
Priyaa , A & Sanjeevi, S 2013, ‘Super resolution mapping of multispectral and hyperspectral images of peechi reservoir, south india’, image, 2010.
Rival, A & Levang, P 2013, ‘La palme de controverse : Palmier à huile et enjeux de développement’, édition Quae. Thenkabail, P. S, Stucky, N, Griscom, B. W, Ashton, M. S, Diels, J, Van Der Meer, B, Enclona,
E 2004, ‘Biomass estimations and carbon stock calculations in the oil palm plantations of African derived savannas using IKONOS data’, International Journal of Remote Sensing, Vol.25, n°23, pp.5447-5472.
